Common Prompting Mistakes

Hey there,

We need to talk about something we’ve seen a lot lately.

Business owners and other professionals saying:

“AI doesn’t work for me.”

Or: 

“I tried ChatGPT, but it gives me generic responses.”

The truth is: AI isn’t useless, you’re just asking it weak questions.
Prompting is a skill. And the better your input, the better your output.

So let’s fix that.

Here are 3 common mistakes we see (and how to fix them):

------------------------------------------------------------------------------
Common Prompting Mistakes

Mistake #1: Being too vague

Vague prompts get vague replies.

For example: “Write a marketing plan.”

This is like walking into a restaurant and saying “bring me food.” What
kind of food? For how many people? Budget?

The Fix: Be specific about the audience, the offer, and the outcome you want.

We like to use the CRAFT template for writing prompts. Not only does it
consistently get great AI responses, it forces us to make sure we’ve covered
all the important details in our prompt.

Here’s the template:

CONTEXT: [Provide overview/background info such as: who you are, company history, overall objectives of the output, relevant environmental information, etc]
ROLE: [Insert the role you want the LLM to play e.g. copywriter with 10+ years
experience in persuasive writing, or content creator specializing in viral
youtube videos]
ACTION: [What do you want AI to do? Providing a list of sequential steps is a great way to ensure the LLM does everything you need.]
FORMAT: [Desired format, e.g. table, paragraph, markdown, code, etc]
TARGET AUDIENCE: [Therecipientof the output. Sometimes this is simply you... but it can be another LLM, your consumer, your boss, your spouse, etc.]
With more specific prompts, you’ll see a dramatic improvement in response
quality.

Mistake #2: Using the wrong model

AI models are evolving fast—and the best model for your use case last
month might not be the best today.

The Fix: Periodically test multiple models.

Use a tool like https://openrouter.ai/ OpenRouter to run the same prompt across different
AI models and compare outputs side-by-side.

You have to buy credits for a small fee, but we find it beats spending time
manually testing 5+ AI models with the same prompt.

Give it a try. You might discover a new favorite you’ve never tried
before, or realize it’s time to switch from the one you’ve been clinging to.

Mistake #3: Not iterating

The situation: You get a meh answer from AI and give up.

The Fix: Treat it like a junior team member. Refine, clarify, and follow up.

One of our favorite techniques is using “response leveling.”

Try saying:

“This is a good level 1 response. Now give me a level 2 response.”
Then you can uplevel further with:
“Take this to a level 3 response with deeper insight and analysis.”

This pushes the model to expand on its original ideas without you having
to completely reframe the prompt or switch models.

Bonus: You can also activate DeepResearch mode (if available).

------------------------------------------------------------------------------
Final Thoughts

If you’re not getting the results you want from AI, the solution isn’t to
give up or invest in pricey tools, it’s better prompting.

Try these three tweaks next time you’re stuck. And let us know if it’s
helpful!

Talk soon,
Brian + Andrea